Scientific Concepts and Research Design
Designing a question that can be answered
A trial exists to answer a question, so the first examinable concepts are the hypothesis and the endpoints that test it. A primary endpoint is the single pre-specified measure that drives sample size and the statistical conclusion; secondary endpoints support or explore, and exploratory endpoints generate hypotheses rather than confirm them. You should be able to classify designs - parallel-group, crossover, factorial, cluster, adaptive, enrichment - and explain why each is chosen. Randomization addresses selection bias and baseline confounding; blinding addresses information bias after allocation. They are separate tools, and questions frequently try to merge them.
Sample size follows from effect size, variability, alpha, power and expected dropout. Alpha is the pre-set Type I error rate (false positive), beta the Type II error rate (false negative), and power is 1 − beta. A p-value is the probability of data at least as extreme as observed, assuming the null hypothesis is true; it is not the probability the null is true, and it is not a measure of effect magnitude. Intention-to-treat preserves randomization and is conservative for superiority; per-protocol answers a narrower question. Non-inferiority designs need an acceptable margin defined in advance and are vulnerable to poor assay sensitivity.
Trap: a scenario describing a well-blinded but non-randomized study, asking which bias is controlled.
Rule of thumb: name the bias, then name the design feature that prevents it - if you cannot pair them, the design claim is wrong. Worked example: a crossover design gives each participant both treatments, so within-subject variability falls and fewer participants are needed, but carryover effects and non-random dropout make it unsuitable for a condition that resolves permanently.